History of OpenAI, the company founded by Musk and Altman behind the ChatGPT phenomenon

December 11, 2015. A few minutes after its registration, a document is uploaded to the Openai.com domain. Three thousand two hundred scarce jokes. They sign it in nine. Among them Sam Altman, Elon Musk and Peter Thiel. But also donor companies such as Amazon and Y-Combinator, the most important startup accelerator in the world. It is the act that kicks off one of the most discussed and fascinating projects of the last decade: OpenAI. The research group from which the ChatGPT phenomenon was born.

That research group, lavishly funded by prominent Silicon Valley profiles, is now a full-fledged company. With shares, a board of directors and investors of weight. It first received funding of one billion from Microsoft. Then, a few days ago, Microsoft bought 40% of it for the same amount.

But OpenAI in 2015 was born as a pure research company. Non-profit. In that document, the founders say they are aware of the enormous potential of artificial intelligence. Both for its ability to make humanity make enormous progress. But also because potentially it could be able to destroy it. To take over the species that has so far dominated planet Earth.

OpenAI, the degree of difference between heaven and hell

Basically there is the awareness that intelligent machines have made enormous strides. It’s no longer about pure computing power. Nor to beat a Russian chess champion in the sport in which he dominated unchallenged. Machines have reached a different evolutionary stage.

In 2015, one felt on the cusp of a revolution whose consequences were difficult to imagine. So it was better to move early. Anticipate the future. Not to limit the development of artificial intelligence – innovation will come sooner or later. But to make it evolve to the fullest.

As we read in one of the most significant passages of the text written seven years ago: “Today, artificial intelligence systems have impressive capabilities. But limit yourself. Yet their limits seem destined to shrink. In the most extreme case, they will soon reach human performance in almost any intellectual task. It is difficult to imagine how much an artificial intelligence with a level equal to or higher than human could benefit society. But it is equally difficult to imagine how much damage it could do if built or used incorrectly”.

Between heaven and hell there seemed to be only one degree of difference. And OpenAI wanted to put itself on that grade.

DeepMind: a project about which little is known, perhaps by choice

OpenAI has become the center from which all kinds of news on the artificial intelligence front have been expected. Often with a mixture of curiosity and hope. It has become the most observed center by experts, together with the other heavyweight of the sector: DeepMind of Alphabet, bought by the Google holding in 2014 for 500 million dollars. Same research focus. Opposite approaches. Two companies that are not only competitors, but at times seem to be on antithetical barricades.

The Alphabet subsidiary doesn’t talk much about the progress of its artificial intelligence. She doesn’t advertise it much. Publish discoveries, yes. Google’s artificial intelligence last June reconstructed and unveiled the structures of 200 million molecules. He also put them in a database accessible to all. It is known that he deals with tools for mathematics. Of techniques for deciphering ancient texts. It is also known that in some cases it has had successes. But what he does is perceived rather coldly. Far away. As if an aura of mystery hung over everything.

Perhaps also an accomplice in the latest story concerning her: the dismissal of one of her employees, Blake Lemoine, after the engineer decided to tell the media that DeepMind’s AI had become conscious. Almost sentient. No confirmation. Indeed, it seems that Lemoine had a sort of inherent fear of his work and its developments that would have led him to imagine something from some ‘actions’ of his AI.

OpenAI: artificial intelligence open to everyone

OpenAI did the opposite. And he did it right away. Everything that came out of the laboratories of the San Francisco headquarters (a historic building in the city where, curiously, Neuralink, Elon Musk’s startup that works on implantable neural interfaces, is also based) was immediately made available. Published, told, disseminated. In the case of the latest version of ChatGPT even made available to everyone. Result? Site inaccessible for days due to too many requests.

In 2016, he put a beta version of a reinforcement machine learning platform online. The year after a software that trains an artificial intelligence through games. Then it was the turn of the robots who learned the techniques of sumo, until they threw the opponent out of the dohyo. Experiments. But it’s ChatGPT that changed everything. It has given everyone the opportunity to understand the real potential of artificial intelligence. But it has also changed the history of the company, which has definitively pulled off the guise of the non-profit born for the good of humanity.

Fears about a bad AI being able to take over

Openai, it has been said, was born with the most recent developments in artificial intelligence. With a basic idea that vibrates like a concern: Musk, Altman, Thiel and the others feared (still fear?) that the great things artificial intelligence is destined to hide a dark side. The fact that those things can be terrible. And since there is no remedy for the development of artificial intelligence, there is no barrier that can stop it, for better or for worse, the only way to create the conditions that can do good and make it open. Publish. Give everyone access. Expand their audience to the highest possible level.

At the beginning Openai programs were even open source, i.e. they could be downloaded and used at will if one is able to work on their code. Then something changed.

Elon Musk left all positions in the research group in 2018, raising the possibility of a conflict of interest between Openai and the company’s artificial intelligence arm of Tesla. The following year Openai becomes a for profit company. No more charitable purposes. No more development done with donation money (Musk remained a major donor). AI is a business. And we need real investments.

OpenAI: the change of pace of 2019, money is needed

In 2019 Openai becomes a limited profit company. Basically, if you put some money into it, you can get a maximum profit of 100 times the invested capital. A few months later Microsoft puts in a billion. From an artificial intelligence laboratory it becomes a company that produces it. It is estimated that $12 million a year would be needed to build its programs for research alone.

The further it went, the more complex the searches became, the more good researchers were needed. And with salaries that are not too unfavorable for economic conditions compared to Google, Amazon or Meta. Microsoft’s money was needed to stay afloat. To “remain relevant”, as stated in the company’s note after the investment of the company founded by Bill Gates.

Critics’ doubts: OpenAI sold its soul for a plate of (very expensive) lentils

In early 2019, Openai announces GPT-2, a language model capable of generating human-level texts. For the researchers, it was a huge leap forward. But their fear was one: they feared it could be used to “spread fake news, spam and disinformation”. But then they decided to share the model, after finding “no strong evidence of misuse”.

Several articles since then have highlighted how the company’s strategy has changed. An article in Boston’s MIT magazine clearly stated that Openai was trying to “capitalize on the panic she herself had helped to spread on AI issues.” In other magazines, such as the American version of Wired, a precise marketing strategy by the company was hypothesized.

Suspicion also raised after the launch in December of the GPT-3.5 chatbot, the version that made ChatGPT pop, today practically inaccessible due to too many access requests arriving from all over the world and which could soon become a paid one. The summary of the critics is one: OpenAI has sold its soul. Not for a plate of lentils.

The company is now aiming for a billion dollar turnover by 2024. A new billion dollar business is being born

On the other hand, it is certain that without a cash flow it is difficult to be able to hire good engineers capable of improving the product. And it is now certain that what Openai and the other companies active in the sector do is what is expected from the development of new technologies. Artificial intelligence seems to have already assumed the scepter of “the next big thing”, the next big innovation to focus on. In terms of development and investment.

After Musk’s exit, today the company is led by Sam Altman. According to Reuters, the company expects a turnover of 200 million in 2023. And one billion for 2024. Today it is valued at 29 billion dollars. The first steps towards being the next giant tech company have already been taken. And the first steps of a new sector. Not for ideas, not even for technology. But for the turnover generated.

After twenty years of absence of great innovations, with the era of the digital economy of social networks a bit on the waning avenue, and in the background the unfulfilled promises of the crypto economy, Silicon Valley needs fresh air to avoid a structural crisis. Artificial is fine too.

Twitter: @arcamasilum